Research On Sentiment Analysis And Product Recommendation Algorithm Based On Topic Feature

With the popularity of the electronic commerce,more and more people choose the Internet to shop,entertain,communicate and work.At the same time,users often need to spend a lot of time on searching and selecting goods or services.How to recommend products and services for users is a problem that the service provider must solve.However,the traditional recommendation algorithm is only a simple analysis of the relationship between the products and the user,not including the sentiment analysis of comment text.To a certain extent,it affects the accuracy of the recommendation results.The thesis recommends products to users accurately from two aspects.That are sentiment analysis and product recommend.First of all,from the perspective of the overall meaning of the Chinese text,we use the topic model LDA to get the topic distribution of the text.In the feature selection and extraction,we consider the topic of short text,except the character of the short-text.We complete the Chinese text sentiment classification.The experimental results show that the accuracy of experiment is improved.The thesis analyzes the user's emotion on the basis of collaborative filtering algorithm and sentiment analysis of Chinese text.By analyzing the emotion of user,reviews,we propose sentiment similarity on the basis of score similarity.Additionally,we propose comprehensive similarity to calculate the score of user to product.Among them,the sentiment similarity calculation method is Topic Feature Reserved Self-training Algorithm the thesis proposed.By comparing with the traditional recommendation algorithm,the experiment proves that the recommendation algorithm based on sentiment analysis has better recommendation effect than the traditional collaborative filtering recommendation algorithm.